Highly-reflective facet-coated multi-wavelength DFB laser array with exact wavelength spacings.

A distributed feedback (DFB) laser array of twenty wavelengths with highly reflective and anti-reflective (HR-AR) coated facets is both theoretically analyzed and experimentally validated. While the HR facet coating enhances high wall-plug efficiency, it inadvertently introduces a random facet grating phase, thereby compromising the lasing wavelength's predictability and the stability of the single-longitudinal-mode (SLM). In this study, two key advancements are introduced: first, the precisely spaced wavelength is achieved with an error of within ±0.2 nm using the reconstruction-equivalent-chirp (REC) technique; second, the random grating phase on the HR-coated facet is compensated by a controllable distributed phase shift through a two-section laser structure. The SLM stability can be improved while the wavelength can be continuously tuned to the standard wavelength grid. The overall chip size is compact with an area of 4000 × 500 µm2. The proposed laser array has a light power intensity above 13 dBm per wavelength, a high side mode suppression ratio above 50 dB, and low relative intensity noise under -160 dB/Hz. These attributes make it apt for deployment in DWDM-based optical communication systems and as a light source for optical I/O.

Photonic convolution accelerator based on a hybrid integrated multi-wavelength laser array by photonic wire bonding for real-time image classification.

We propose and experimentally demonstrate a compact and efficient photonic convolution accelerator based on a hybrid integrated multi-wavelength DFB laser array by photonic wire bonding. The photonic convolution accelerator operates at 60.12 GOPS for one 3 × 3 kernel with a convolution window vertical sliding stride of 1 and generates 500 images of real-time image classification. Furthermore, real-time image classification on the MNIST database of handwritten digits with a prediction accuracy of 93.86% is achieved. This work provides a novel, to the best of our knowledge, compact hybrid integration platform to realize the optical convolutional neural networks.

Silicon-Nanomembrane-Based Broadband Synaptic Phototransistors for Neuromorphic Vision.

Neuromorphic vision has been attracting much attention due to its advantages over conventional machine vision (e.g., lower data redundancy and lower power consumption). Here we develop synaptic phototransistors based on the silicon nanomembrane (Si NM), which are coupled with lead sulfide quantum dots (PbS QDs) and poly(3-hexylthiophene) (P3HT) to form a heterostructure with distinct photogating. Synaptic phototransistors with optical stimulation have outstanding synaptic functionalities ranging from ultraviolet (UV) to near-infrared (NIR). The broadband synaptic functionalities enable an array of synaptic phototransistors to achieve the perception of brightness and color. In addition, an array of synaptic phototransistors is capable of simultaneous sensing, processing, and memory, which well mimics human vision.

Photonic convolutional neural network with robustness against wavelength deviations.

We experimentally explore the practicality of integrated multiwavelength laser arrays (MLAs) for photonic convolutional neural network (PCNN). MLAs represent excellent performance for PCNN, except for imperfect wavelength spacings due to fabrication variation. Therefore, the performance of PCNN with non-ideal wavelength spacing is investigated experimentally and numerically for the first time. The results show that there exists a certain tolerance for wavelength deviation on the degradation of the structural information of the extracted feature map, leading to the robustness of photonic recognition accuracy under non-ideal wavelength spacing. The results suggest that scalable MLAs could serve as an alternative source for the PCNN, to support low-cost optical computing scenarios. For a benchmark classification task of MNIST handwritten digits, the photonic prediction accuracy of 91.2% for stride 1 × 1 scheme using the testing dataset are experimentally obtained at speeds on the order of tera operations per second, compared to 94.14% on computer. The robust performance, flexible spectral control, low cost, large bandwidth and parallel processing capability of the PCNN driven by scalable MLAs may broaden the application possibilities of photonic neural networks in next generation data computing applications.

10×10 Gbs directly modulated DFB laser array based on the REC technique.

We proposed and experimentally demonstrated a directly modulated distributed feedback (DFB) laser array with a transmission rate of 100 Gbps (10c h a n n e l s×10G b p s). The grating design is based on the reconstruction equivalent chirp (REC) technique, which enables precise control of the channel wavelength spacing to 100 GHz, as specified in the ITU-DWDM standard. DFB laser arrays incorporating the REC technique demonstrate excellent consistency performance, with a side-mode suppression ratio exceeding 48 dB, threshold current of approximately 20 mA, and modulation bandwidth of greater than 13 GHz at a bias current of 100 mA. We evaluated the laser's performance by loading a 10 Gbps nonreturn-to-zero signal onto the laser using direct modulation and transmitting it over a 10 km single-mode fiber. Based on our experimental results, the proposed DFB laser array is promising to be utilized in the next generation of low-cost, 100 Gbps DWDM communication systems.

High-power tapered two-section distributed feedback laser based on a chirped sampled grating.

We propose and theoretically study a two-section high-power distributed feedback (DFB) laser with three equivalent phase shifts (3EPSs). A tapered waveguide with a chirped sampled grating is introduced to amplify the output power and keep a stable single-mode operation. The simulation exhibits the maximum output power and side mode suppression ratio of a 1200 µm length two-section DFB laser as high as 306.5 mW and 40 dB, respectively. Compared with traditional DFB lasers, the proposed laser has a higher output power, which may benefit wavelength division multiplexing transmission systems, gas sensors, and large-scale silicon photonics.

Experimental demonstration of a photonic convolutional accelerator based on a monolithically integrated multi-wavelength distributed feedback laser.

We propose and experimentally demonstrate a simple and energy-efficient photonic convolutional accelerator based on a monolithically integrated multi-wavelength distributed feedback semiconductor laser using the superimposed sampled Bragg grating structure. The photonic convolutional accelerator operates at 44.48 GOPS for one 2 × 2 kernel with a convolutional window vertical sliding stride of 2 and generates 100 images of real-time recognition. Furthermore, a real-time recognition task on the MNIST database of handwritten digits with a prediction accuracy of 84% is achieved. This work provides a compact and low-cost way to realize photonic convolutional neural networks.

Optically Stimulated Synaptic Devices Based on the Hybrid Structure of Silicon Nanomembrane and Perovskite.

Optoelectronic synaptic devices have been attracting increasing attention due to their critical role in the development of neuromorphic computing based on optoelectronic integration. Here we start with silicon nanomembrane (Si NM) to fabricate optoelectronic synaptic devices. Organolead halide perovskite (MAPbI3) is exploited to form a hybrid structure with Si NM. We demonstrate that synaptic transistors based on the hybrid structure are very sensitive to optical stimulation with low energy consumption. Synaptic functionalities such as excitatory post-synaptic current (EPSC), paired-pulse facilitation, and transition from short-term memory to long-term memory (LTM) are all successfully mimicked by using these optically stimulated synaptic transistors. The backgate-enabled tunability of the EPSC of these devices further leads to the LTM-based mimicking of visual learning and memory processes under different mood states. This work contributes to the development of Si-based optoelectronic synaptic devices for neuromorphic computing.

On-chip mode converter based on two cascaded Bragg gratings.

We propose an on-chip mode converter via two cascaded Bragg reflection processes. A forward conversion between two guided modes can be achieved with the aid of an additional mode. The proposed structure is theoretically studied and simulated via the rigorous three-dimensional finite-difference time-domain (3D-FDTD) method. The bandwidth and central wavelength of the proposed mode converter can be adjusted according to our theoretical analysis and simulation results. By applying the similar design approaches as fiber Bragg gratings, conversion spectra with different shapes can be obtained. As an example, several mode converters with bandpass and sidelobe-reduced spectra are designed. We also investigate and verify the mode conversion by experiment. Therefore, the proposed method may pave a new path for the mode converters with desired conversion spectra.

Study on sampled waveguide grating with anti-symmetric periodic structure.

An anti-symmetrically sampled Bragg grating (ASBG) with single mode waveguide is proposed and investigated for the first time. Based on anti-symmetric periodic structure, the coupling coefficient between the forward and backward guided modes becomes zero, thus nearly no light is reflected. Besides, the equivalent tilted grating effect with radiation mode coupling is found. If another anti-symmetrically sampling structure is imposed to form a sampled grating, the 0th sub-grating can be avoided, while the ± 1st sub-gratings are adjusted as uniform gratings with normal performances. This will be very benefit for some special applications such as distributed feedback (DFB) lasers based on Reconstruction-equivalent-chirp (REC) technique where 0th order resonance can be avoided. In addition, error analysis for the proposed structure is also performed for practical applications.